摘要
水声信号的局部可预测性在水声信号处理中具有重要作用,它是解决非平稳信号检测问题的基础。基于非线性时间序列局部可预测性原理,采用人工神经网络技术,研究了水声信号的神经网络预测,讨论了预测模型的建立和网络结构参数的设计。分别采用BP网络和RBF网络对实际舰船水声信号进行预测,通过对仿真数据和实际舰船辐射噪声数据预测的结果分析,得出了两种网络预测模型的误差分布,提出了减小预测误差的有效方法。为今后进一步开展水声信号预测研究奠定了基础。
Local predictability is important in underwater acoustic signal processing, which is the basis in detection of non-stationary signals. Based on prediction of nonlinear time series and artificial neural networks, prediction of underwater acoustic signals is studied in this paper. Prediction models and design of network parameters are discussed. Predictions of both simulated data and real ship radiated noise data are made using BP and RBF network. Error distribution of the two kinds of network prediction models is obtained, and a method for effective reduction of prediction error is presented. The results are useful in the further study of underwater acoustic signal processing.
出处
《声学技术》
CSCD
北大核心
2006年第3期226-229,共4页
Technical Acoustics
基金
国家自然科学基金资助项目(104740179)
关键词
水声信号
神经网络
预测
underwater acoustic signal
neural network
prediction
